Table of Contents
The use of AI in product development is revolutionizing the way companies drive innovation and meet user needs. In the face of these massive changes, today’s technology leaders must ensure their efforts have real impact and aren’t just empty words.
A solid foundation consists of a clear understanding of customer needs, market dynamics and the technology landscape. This requires a balanced mix of user-centricity, purposefully limiting AI optimization to relevant areas and ensuring that the underlying requirements and processes are ready for the integration of AI tools. This way, organizations can create innovative, sustainable and user-friendly solutions that unlock the true value of AI in product innovation.
VP of Engineering at Box.
Focus on your users
AI is all the rage right now, but unlike recent hype cycles, the natural language and reasoning capabilities of large language models (LLMs) have the potential for broad impact across a wide range of products. But adding AI to a well-functioning product just to be trendy will only confuse users.
Before implementing AI, companies should ask themselves whether AI is actually addressing problems or gaps. User evaluation is essential. How can AI improve the employee experience without losing sight of the product strategy? Interactions that would benefit from a natural language interface, or manual workflows that could be streamlined, can also be important.
A well-known example: When searching on the online platform AirBnB, users filter options based on criteria such as price or number of bedrooms. Simply replacing a simple filter with users typing a question in natural language does not unlock new possibilities and creates a more cumbersome flow with a higher risk of unexpected results. Yet, even with multiple filters, it is not always easy to find what you are looking for. Modeling the long tail of personal criteria as effective filters is difficult. AI’s ability to understand the nuances of natural language can make all the difference. With AI-driven search, there are no limits to personalization. At the same time, fast, meaningful and functional filters should not be sacrificed.
While it’s relatively easy to create a compelling demo using new AI technologies, it’s challenging to build a useful product. Building a step-by-step process allows you to learn from user feedback and is crucial to creating a valuable and compelling experience.
Beware of premature fine-tuning!
One of the exciting aspects of this latest wave of AI is its ability to become hyper-personalized. What once required human nuance and intent can now be digitized and made accessible at scale.
The fascination with technology should not get in the way of knowing everything about practical product development. While fine-tuning a custom model may seem tempting, it is a form of premature optimization that locks in a series of choices before the right product fit is found. Prematurely fine-tuning an AI model slows down the iteration speed and increases maintenance costs, ultimately hindering the speed of innovation.
So how do you create a customized experience? It’s all about the prompt. The prompt is a great place to set the tone for the interaction — trust, cultural or industry fit, brand voice, and more. It should communicate any proprietary information the model will need to use. The prompt should also summarize the context that new hires need to be provided to complete the task.
This approach provides the flexibility to improve and adapt incrementally as both the underlying technology and the understanding of how to use it evolve. The sophistication of task structuring ultimately becomes a key differentiator for a product. AI models are like black boxes – a query leads to an answer. Even small changes can lead to dramatic changes in quality. Early implementation of a quality assurance process enables effective evaluation of improvements and early detection of degradation.
The basis for rapid innovation
To keep up with the pace of change in AI, a team needs to be able to evolve quickly. A solid foundation starts with building an AI platform that paves the way for developers and enables both rapid iteration and consistency across the product. Consideration should also be given to standardization on approved vendors and models, a basic query framework, an approach to quality testing, and basic patterns and functions for extracting relevant data from common data sources to serve as context in the query.
While there can be many challenges in simplifying an AI platform, there should not be too much emphasis on centralization. It’s not about the technology, but how it’s integrated into the product. Teams responsible for a specific aspect of a product are best placed to identify and optimize suitable use cases. Therefore, all members of a product development team must be able to successfully use AI in their respective areas.
We provide an overview of the best productivity tools.
This article was produced as part of TechRadarPro’s Expert Insights channel, where we showcase the best and brightest minds in the technology sector today. The views expressed here are those of the author and do not necessarily represent those of TechRadarPro or Future plc. If you’re interested in contributing, you can read more here:
